AI技术助力基础设施预测性维护 提前数周预警故障

[来源: CP] [分类提示: NEWS] 标题: AI Could Predict Infrastructure Failures Weeks in Advance AI-powered analytics are helping infrastructure engineers move from reacting to failures toward predicting problems before they become emergencies, according to Vahid Abdollahi, an Applied AI Scientist at Bentley Systems. Traditional monitoring often alerts operators only after a measurement crosses a fixed threshold. AI, however, can analyze data from multiple sensors to identify unusual patterns and potentially provide four to six weeks of advance warning of dangerous conditions. The process begins with reliable sensor data. AI systems must first identify faulty, drifting, or abnormal readings before analyzing genuine structural changes. Techniques such as Seasonal-Trend Decomposition can then separate normal seasonal behavior from unexplained changes that may indicate structural problems. By analyzing relationships between sensors, AI can also distinguish equipment faults from real events. For example, changes in rainfall, ground pressure, and movement can reveal potential problems such as internal erosion before significant movement occurs. Vahid Abdollahi, Applied AI Scientist, Bentley Systems Abdollahi emphasizes that AI is designed to support engineers, not replace them. Feature-importance analysis can show why a model has identified a risk, while generative AI can turn complex sensor data into clear, prioritized recommendations for engineers. Bentley Systems’ iTwin IoT brings these capabilities togethe